Spironolactone alleviates schizophrenia-related reversal learning in Tcf4 transgenic mice subjected to social defeat

Abstract

Cognitive deficits are a hallmark of schizophrenia, for which no convincing pharmacological treatment option is currently available. Here, we tested spironolactone as a repurposed compound in Tcf4 transgenic mice subjected to psychosocial stress. In this '2-hit' gene by environment mouse (GxE) model, the animals showed schizophrenia-related cognitive deficits. We had previously shown that spironolactone ameliorates working memory deficits and hyperactivity in a mouse model of cortical excitatory/inhibitory (E/I) dysbalance caused by an overactive NRG1-ERBB4 signaling pathway. In an add-on clinical study design, we used spironolactone as adjuvant medication to the standard antipsychotic drug aripiprazole. We characterized the compound effects using our previously established Platform for Systematic Semi-Automated Behavioral and Cognitive Profiling (PsyCoP). PsyCoP is a widely applicable analysis pipeline based on the Research Domain Criteria (RDoC) framework aiming at facilitating translation into the clinic. In addition, we use dimensional reduction to analyze and visualize overall treatment effect profiles. We found that spironolactone and aripiprazole improve deficits of several cognitive domains in Tcf4tg x SD mice but partially interfere with each other's effect in the combination therapy. A similar interaction was detected for the modulation of novelty-induced activity. In addition to its strong activity-dampening effects, we found an increase in negative valence measures as a side effect of aripiprazole treatment in mice. We suggest that repurposed drug candidates should first be tested in an adequate preclinical setting before initiating clinical trials. In addition, a more specific and effective NRG1-ERBB4 pathway inhibitor or more potent E/I balancing drug might enhance the ameliorating effect on cognition even further.

Fig. 1. Experimental layout and data flow.

A Experimental timeline. Tcf4tg-mice were all subjected to social stress during early adolescence (Tcf4tgSD). The behavioral tests were performed according to the previously published PsyCoP protocol with tail suspension test and pre-pulse inhibition test reversed and in order of increasing aversiveness. B Overview of the data analysis pipeline. Data acquisition and preprocessing were semi-automated. All data were collected in a RData dump for further analysis in FlowR. The variables were categorized in domains following the Research Domain Criteria (RDoC) framework. Canonical discriminant analysis (CDA) was used for dimension reduction. The resulting neurocognitive profiles were visualized in dimension plots of the dimension reduction and a heatmap of the weights of each variable in the CDA result.

Fig. 2. Adjuvant spironolactone treatment interferes with the beneficial effect of aripiprazole on serial reversal learning performance.

Variables measuring aspects of the cognition domain are shown in panels (A–F). A The rate of spontaneous alternations in the Y-maze test (YM) was used as an indicator of working memory performance. B The number of trials a subject needs to reach the learning criterion in the sequential probability ratio test (SPRT) reflects the learning speed, with lower values indicated higher learning performance. The graph shows group means of that learning speed with the standard error of the mean (SEM). C The performance in the first reversal indicates learning flexibility following spatial learning in the place learning phase. D Performance in the more demanding serial reversal learning task was quantified as area under the learning curve (AUC), computed as sum of the rolling mean across all reversal phases. E, F Freezing behavior in the cued (E) and contextual (F) fear conditioning tasks. G–J Prepulse inhibition results. Included are (G) baseline startle and prepulse inhibition at sound levels of H 70 dBA, I 75 dBA, and J 80 dBA. Data are shown as box plots with whiskers extending to no more than 1.5-fold IQR; *p < 0.05, **p < 0.01, ***p < 0.001, n.s. not significant; p-values are FDR-adjusted and refer to Wilk’s lambda testing two-way ANOVA; n = 19/19/30/30; Plc placebo, Spl spironolactone treatment, Apz aripiprazole treatment, S spironolactone term, A aripiprazole term, SxA interaction term, P-S spironolactone effect in placebo-treated mice, A-S spironolactone effect in aripiprazole-treated mice.

Fig. 3. Aripiprazole treatment increases measures of negative valence in Tcf4SD mice.

All three variables associated with the positive valence domain. A Sucrose preference quantified in an IntelliCage task, measured as preference score from −1 (absolute avoidance) to 1 (exclusive preference). B The same preference score representing preference for the rewarded corner in a spatial positive reinforcement-learning paradigm in the IntelliCage (place learning). C Path choice quantified as rotation rate in the open field test was interpreted as nervous or hectic behavior. D Center time in the open field test and E baseline freezing behavior in the fear conditioning test were used as proxies of anxiety level. F Time immobile in the tail suspension test quantifying the absence of struggling behavior. Data are shown as box plots with whiskers extending to no more than 1.5-fold IQR; *p < 0.05, **p < 0.01, ***p < 0.001, n.s. not significant; p-values are FDR-adjusted and refer to Wilk’s lambda testing two-way ANOVA; n = 19/19/30/30; Plc placebo, Sp spironolactone treatment, Apz aripiprazole treatment, S spironolactone term, A aripiprazole term, SxA interaction term, P-S spironolactone effect in placebo-treated mice, A-S spironolactone effect in aripiprazole-treated mice.

Fig. 4. Aripiprazole reduces measures of arousal in Tcf4SD mice.

A General locomotor activity level in the IntelliCage (IC) was quantified as instantaneous visit frequency. B In addition, the nocturnality of activity in the IC was measured with a preference (nocturnality) score similar to sucrose preference, where −1 indicates exclusive daytime activity and +1 only nighttime activity. C, D Moreover, spontaneous, novelty-induced activity was assessed as (C) total number of choices in the Y-maze test (YM) and (D) mean speed in the open field test (OF). E, F Circadian activity distribution in 1-h bins, quantified as instantaneous frequency. The data is shown as group (line) and its standard error (ribbon). Data are shown as box plots with whiskers extending to no more than 1.5-fold IQR; *p < 0.05, **p < 0.01, ***p < 0.001, n.s. not significant; p-values are FDR-adjusted and refer to Wilk’s lambda testing two-way ANOVA; n = 19/19/30/30; Plc placebo, Spl spironolactone treatment, Apz aripiprazole treatment, S spironolactone term, A aripiprazole term, SxA interaction term, P-S spironolactone effect in placebo-treated mice, A-S spironolactone effect in aripiprazole-treated mice.

Fig. 5. Canonical discriminant analysis reveals a clear interaction between aripiprazole and spironolactone treatment.

A The dimension plot of the first two canonical components shows a clear separation of aripiprazole-treated groups from their respective placebo-treated references as well as the Plc-Spl from the Plc-Plc group. The first canonical component (Can1) explains 79.3% of the total canonical correlation, while the second component (Can2) only explains 14.3%. B The weights of the single variables (canonical coefficients) are plotted on top of the data ellipses, showing which variables drive group separation in this analysis. C The heatmap shows the difference of the mean Z-score of each treatment group compared to placebo control. The hierarchical clustering on top of the Z-score profiles supports the stronger and less specific effect of aripiprazole (blue colors) on the behavioral phenotype compared to spironolactone (dark shades). D In a multivariate ANOVA, we found an interaction of spironolactone and aripiprazole treatment (SxA: F(19,67) = 4.52, p = 2.09 × 10⁻⁶) with significant spironolactone effects in both placebo and aripiprazole-treated mice (P-S: F(19, 12) = 4.55, p = 5.07 × 10⁻³; A-S: F(19, 37) = 2.40, p = 0.0114). When looking at the CDA results from individual terms of the multivariate linear model and their corresponding simple-effects models, we find several strong contributions to the interaction of aripiprazole and spironolactone treatment. In contrast to the collapsed factor model shown in A and B, this analysis deconvolutes the two factors and their interaction strictly. The corresponding simple-effects models suggest that the direction of spironolactone’s effect depends on aripiprazole treatment, having positive weights in the placebo group’s canonical component and negative weights in the aripiprazole-treated groups. *p < 0.05, **p < 0.01, ***p < 0.001, n.s. not significant; p-values are FDR-adjusted and refer to Wilk’s lambda testing in a multivariate two-way ANOVA with subsequent univariate two-way ANOVAs; In case of a statistically significant interaction, the spironolactone main effect was tested for placebo and aripiprazole treatment in simple-effects ANOVAs; n = 19/19/30/30; Plc placebo, Spl spironolactone treatment, Apz aripiprazole treatment, S spironolactone term, A aripiprazole term, SxA interaction term, P-S spironolactone effect in placebo-treated mice, A-S spironolactone effect in aripiprazole-treated mice.